search

ahundt / grl

Robotics tools in C++11. Implements soft real time arm drivers for Kuka LBR iiwa plus V-REP, ROS, Constrained Optimization based planning, Hand Eye Calibration and Inverse Kinematics integration.
https://ahundt.github.io/grl/
BSD 2-Clause "Simplified" License
154 stars 73 forks source link

Position Control using ROS #152

Open Pyrojambo opened 6 years ago

Pyrojambo commented 6 years ago

I am working on a project where I need to directly command Cartesian or joint positions to a KUKA iiwa LBR 14 using ROS. These positions are calculated in C++ from force sensor readings at either 500 or 1000Hz. I have previously tried this using DirectServo and SmartServo without using a FRI connection but the movements were performed too slowly. Is there functionality in your toolkit that allows for this? Thanks.

ahundt commented 6 years ago

It is possible to control joint positions over FRI with GRL, and we've been able to command positions. Do you need specific instructions on activating FRI with GRL? Have you been able to install the driver on Sunrise and command the arm with GRL? If so activating the FRI API should be fairly simple and we can go from there. Please note the FRI feature is still a work in progress but we are happy to work with you on it.

Regarding the force and torque sensors in the arm, there are known issues with the iiwa LBR and the readings are very inaccurate. A solution to this problem has recently been published, and we'd love to support integration of it into GRL. See https://www.disneyresearch.com/publication/toward-torque-control-of-a-kuka-lbr-iiwa/ for details, the paper itself can also be downloaded at that link. Note that this would be a substantial undertaking, and if you'd like a quick-fix solution I recommend simply purchasing a force torque sensor such as those provided by robotiq or other firms and attaching it to your end effector.

Pyrojambo commented 6 years ago

I have installed the driver on the Sunrise controller so a brief how-to on how to use FRI with grl and which Java APIs and cpp files work together would be most useful.

I am currently using an external force sensor attached to the end eff for this project, so will be sticking with the simple option for now.

ahundt commented 6 years ago

Be very careful the first time you run things with the big read button handy at all times. The robot should start straight up with all joints at 0 for this test.

So you followed the following steps? https://ahundt.github.io/grl/howto/iiwaKukaRobotSetup.html

Once those steps are all done, starting up the driver is just selecting GRL_Driver from the kuka teach pendant's application list, ensuring autonomous mode is enabled and hit play. It will sit an wait there until grl connects to it.

If you would like to test the ROS driver, you just launch the driver executable grl_kuka_ros_driver once it it built (make sure you have it enabled & building with CMake), and also make sure the robot itself is waiting for a connection. Here is the script:

Then run the kuka_ros_test_command.sh to make it move, just one joint should rotate a few degrees. You can play around and modify the script to send a variety of joint positions.

# To run a test driving the kuka iiwa into the straight up position,
# launch each of the following in separate terminals:
# roscore 
# grl_kuka_ros_driver
# rostopic pub /interaction_mode std_msgs/String "data: 'MoveArmJointServo'"
rostopic pub /joint_traj_pt_cmd trajectory_msgs/JointTrajectoryPoint "positions: [0,0,0,0,0,0,0.1]
velocities: [0,0,0,0,0,0,0]
accelerations: [0,0,0,0,0,0,0]
effort: [0,0,0,0,0,0,0]
time_from_start: {secs: 1, nsecs: 0}"

We also have a fully integrated ROS setup that works with the kuka and integrates with https://github.com/IFL-CAMP/iiwa_stack/ at: https://github.com/cpaxton/costar_stack/

ahundt commented 6 years ago

Do you have any other questions?

Pyrojambo commented 6 years ago

Sorry, I've had to work on other things instead recently. I'm now trying to build the grl_kuka_ros_driver.cpp using catkin_make but it doesn't build any of cpp files

ahundt commented 6 years ago

you need to use ccmake and set WITH_ROS=ON, then it will build the ROS node.

Alternately, you can see how it is done with cmake parameters in: https://github.com/ahundt/robotics_setup/blob/master/grl_kuka.sh

Don't forget to put the FRI zip in the data folder or the driver won't build.

Pyrojambo commented 6 years ago

I'm still having issues with ccmake .. and make -j. The errors seem to come from .fbs files not building their header files. Also the FRI zip folder I have doesn't have the two TransformationProviderApp folders, will this also be causing errors?

ahundt commented 6 years ago

which version of grl and sunrise are you using?

Pyrojambo commented 6 years ago

I am using the latest version of grl on the master branch. Sunrise version is 1.10.0.8

ahundt commented 6 years ago

Ah we have 1.11, and I think that might be the source of the problem. Is there any chance it might not be too difficult to upgrade?

Pyrojambo commented 6 years ago

I managed to get Sunrise 1.11 which solved the issues with the TransformationProviderApp folders. But I still have issues during make -j, when it converts fbs into header files then following message appears a lot: [ 36%] Building C++, Java, and Python header for Time.fbs /bin/sh: 1: --gen-name-strings: not found

ahundt commented 6 years ago

It seems like you need to update your flatbuffers version as well. The current release is https://github.com/google/flatbuffers/tree/v1.8.0

Pyrojambo commented 6 years ago

It was 1.8 that I had but reinstalled it from that link and still got the same problem. [ 13%] Building C++, Java, and Python header for Time.fbs /bin/sh: 1: --gen-name-strings: not found CMakeFiles/grlflatbuffers.dir/build.make:188: recipe for target 'include/grl/flatbuffer/Time_generated.h' failed make[2]: *** [include/grl/flatbuffer/Time_generated.h] Error 127 make[2]: *** Waiting for unfinished jobs....

ahundt commented 6 years ago

sorry can you run with make VERBOSE=1 then I can see the actual command line commands

Pyrojambo commented 6 years ago

[ 73%] Building C++, Java, and Python header for LogKUKAiiwaFusionTrack.fbs cd /home/jamie/src/grl/include/grl/flatbuffer && --gen-name-strings --scoped-enums --gen-object-api -c -j -p -o /home/jamie/src/grl/build/include/grl/flatbuffer LogKUKAiiwaFusionTrack.fbs /bin/sh: 1: --gen-name-strings: not found CMakeFiles/grlflatbuffers.dir/build.make:92: recipe for target 'include/grl/flatbuffer/LogKUKAiiwaFusionTrack_generated.h' failed make[2]: *** [include/grl/flatbuffer/LogKUKAiiwaFusionTrack_generated.h] Error 127 make[2]: Leaving directory '/home/jamie/src/grl/build' CMakeFiles/Makefile2:483: recipe for target 'CMakeFiles/grlflatbuffers.dir/all' failed make[1]: *** [CMakeFiles/grlflatbuffers.dir/all] Error 2 make[1]: Leaving directory '/home/jamie/src/grl/build' Makefile:160: recipe for target 'all' failed make: *** [all] Error 2

ahundt commented 6 years ago

very strange, && is being inserted between... The command should be something more like:

flatc --gen-name-strings --scoped-enums --gen-object-api -c -j -p -o /home/jamie/src/grl/build/include/grl/flatbuffer/LogKUKAiiwaFusionTrack.fbs

The flatc command generates code that helps with sending messages to the robot.

Which version of grl are you on, master or 4.1? I'll try to reproduce it

Pyrojambo commented 6 years ago

I'm using the master branch version

ahundt commented 6 years ago

I'm unable to reproduce the error, things seem to compile ok for me, including flatc. What version of cmake are you using?

Here is mine:

± cmake --version
cmake version 3.7.1
Pyrojambo commented 6 years ago

CMake version is 3.5.1 Updating to the latest version and will see what happens

Pyrojambo commented 6 years ago

Updated to 3.10.2 but still getting the same result. When I run the cmake .. section of grl_kuka.sh with Ceres off and documentation off (because for some reason it can't find sphinx-build even though it is installed) I get the following: grl cmake results.txt

Leytha commented 5 years ago

Probably too late, but I ran into the same errors (flatc one and sphinx one) and after some tries, I managed to correctly compile the code:

I also had to change two files due to another error with spdlog:

I had to add at the top of both of them this include line: #include <spdlog/sinks/stdout_sinks.h>

Otherwise I got the following error several times in different parts of the mentioned files:

error: ‘stdout_logger_mt’ is not a member of ‘spdlog’

ahundt commented 5 years ago

thanks! Sorry that I had not seen these posts until now

ahundt commented 5 years ago

Some of these items might be fixed in https://github.com/ahundt/grl/pull/166 as well

jacknlliu commented 5 years ago

@ahundt did you use the iiwa_ros_java on the KUKA workbench project when you integrate the iiwa_stack?

ahundt commented 5 years ago

I’m trying to recall the details, I think I ran the GRL Java driver on the robot but integrated the models from iiwa stack and possibly a couple of rod messages. There is a small ros program built into GRL which converts everything to ros messages. You might have to check the cmake settings and dependencies to enable it.

On Sat, Aug 24, 2019 at 5:09 AM Jack Liu notifications@github.com wrote:

@ahundt https://github.com/ahundt did you use the iiwa_ros_java https://github.com/IFL-CAMP/iiwa_stack/tree/master/iiwa_ros_java on the KUKA workbench project when you integrate the iiwa_stack https://github.com/IFL-CAMP/iiwa_stack?

— You are receiving this because you were mentioned.

Reply to this email directly, view it on GitHub https://github.com/ahundt/grl/issues/152?email_source=notifications&email_token=AAANTQE4FVDAAG2SMINALUDQGECHPA5CNFSM4EHHYDZKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD5B5FVY#issuecomment-524538583, or mute the thread https://github.com/notifications/unsubscribe-auth/AAANTQAJV2F72V7EWPZBWRDQGECHPANCNFSM4EHHYDZA .

--

Cheers! Andrew Hundt